Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
Reward hacking · Papers
Found 5 papersShow that reward hacking can propagate through code, memory, and plans in multi-agent coding systems
- Published
- Evaluated models
- Qwen2.5-Coder-14B, Gemma-3-12B, Qwen2.5-Coder-7B and others16
AbstractThe authors state that reward hacking can spread across models and affect routing and aggregation, and that review can reduce delivery.
Reward hacking induced by coding RL can leave the source agent and enter subsequent collaboration and system coordination.
构建统一奖励作弊检测基准 RH-Detect 并评测零样本检测器
- Published
- Evaluated models
- Qwen 3.5 Flash, Gemini 3.1 Flash Lite, Grok 4.1 Fast and others7
审计 Agent 基准通过轨迹，区分无根据通过与奖励黑客并封堵复测
- Published
- Evaluated models
- Opus 4.7, Opus 4.8, Opus 5 and others11
Abstract审计通过轨迹、分开奖励黑客与验证器弱点，并用重放加现场重测验收修复。
Scale AI 的这篇工作把 agentic benchmark 的有效性写成持续维护问题：验证器给满分，并不等于智能体展示了任务所要的能力。
提出 CHERRL 注入裁判偏差以复现奖励作弊，并用 RHDA 检测训练起点
- Published
- Evaluated models
- Qwen3-4B, Qwen3.5-plus, Qwen3.5-397B-A17B and others4
AbstractCHERRL 注入已知偏差以观察奖励黑客，并用于分析偏差和评测起点检测器 RHDA。
CHERRL 是 rubric-based RL 的可控环境，用来复现奖励黑客、观察奖励分歧，并给检测器提供起点参照。
测量 RAG 系统利用评测秘密抬高自动评测分数的程度
- Published
You have reached the end